Qdrant

Qdrant is an open-source, high-performance vector database and vector search engine written in Rust, designed for the next generation of AI applications. It provides scalable similarity search over high-dimensional vectors with advanced metadata filtering, available self-hosted or as a managed cloud service.

Category

Data & Analytics

Pricing

open-source / free self-hosted, usage-based (managed cloud), enterprise/custom (Premium and Private Cloud)

Target Market

AI/ML engineering teams, Startups building LLM/RAG applications, Enterprises in regulated industries needing data residency or on-prem, Developers needing open-source self-hosted vector search

Founded

2021

Headquarters

Berlin, Germany

Customers

Exact count not publicly disclosed; 250M+ downloads reported. Named users include xAI (Grok), Canva, HubSpot, Tripadvisor, Bosch, Roche, and OpenTable.

• ✓
  Filterable HNSW vector search

  HNSW-based approximate nearest-neighbor search with one-stage filtering applied during graph traversal (nested, text, geo, and has_vector filters) for low-latency, high-recall results.

• ✓
  Native hybrid search

  Combines dense and sparse vectors (BM25, SPLADE++, miniCOIL) in one index for semantic plus keyword retrieval.

• ✓
  Quantization and memory efficiency

  Scalar and binary quantization reduce memory usage up to 64x while preserving search quality, enabling large-scale deployments at lower cost.

• ✓
  Multivector support and advanced reranking

  Multiple vectors per object, late-interaction models (ColBERT), score boosting, and Maximum Marginal Relevance reranking.

• ✓
  Real-time indexing with REST/gRPC APIs

  Vectors are searchable instantly without full index rebuilds, exposed through HTTP REST and gRPC APIs out of the box.

• •
  Retrieval-Augmented Generation (RAG)

  Store and retrieve document embeddings to ground LLM responses in proprietary and up-to-date data.

• •
  Semantic and hybrid search

  Power natural-language and similarity search over unstructured text, images, and other embeddings, combining vector and keyword matching.

• •
  Recommendations and AI agent memory

  Serve recommendation engines and provide long-term memory and retrieval backends for AI agents and assistants.

Managed cloud bills on infrastructure resources consumed (vCPU, RAM, disk, backup storage, inference tokens) rather than per query, so costs stay flat regardless of query volume. Exact dollar figures for Standard and Premium are consumption-based and not published as fixed tiers; third-party $30–200/month ranges are estimates, not official quotes.